flight speeds and climb rates of brent geese: mass-dependent differences between spring and autumn...

JOURNAL OF AVIAN BIOLOGY 31: 215–225. Copenhagen 2000

Flight speeds and climb rates of Brent Geese: mass-dependentdifferences between spring and autumn migration

Martin Green and Thomas Alerstam

Green, M. and Alerstam, T. 2000. Flight speeds and climb rates of Brent Geese:mass-dependent differences between spring and autumn migration. – J. Avian Biol.31: 215–225.

Aerodynamic theories of bird flight predict that horizontal flight speed will increasewith increasing load whereas vertical flight speed will decrease. Horizontal flightspeed for birds minimizing overall time on migration is predicted to be higher thanflight speed for birds minimizing energy expenditure. In this study we compare flightspeeds of Brent Geese Branta b. bernicla recorded by tracking radar and optical rangefinder during spring and autumn migration in southernmost Sweden, testing theabove-mentioned predictions. Geese passing Sweden in spring are substantiallyheavier than in autumn and there might also be a stronger element of time-selectionin spring than in autumn. Recorded airspeeds were significantly higher in spring(mean 19.0 m s−1) than in autumn (mean 17.3 m s−1), the average difference beingslightly larger than predicted due to the mass difference alone. The effects on airspeedof wind, vertical speed, flock size and altitude were also analysed, but none of thesefactors could explain the seasonal difference in airspeed. Hence, the results supportthe hypothesis of mass-dependent flight speed adjustment. The difference between thetwo seasons was not large enough to corroborate the hypothesis of a stronger elementof time-selection in spring, but this hypothesis cannot be rejected. Vertical flightspeeds were lower in spring than in autumn, supporting a negative effect of load onbirds’ flight power margin.

M. Green and T. Alerstam, Department of Animal Ecology, Lund Uni6ersity, EcologyBuilding, S-223 62 Lund, Sweden. E-mail: [email protected]

On theoretical grounds birds are expected to adjusttheir flight speed in relation to a multitude of factors,e.g. depending on whether their flight is adapted tominimize power, energy cost per distance covered ortotal duration of the migratory journey. Flight speedwill also be influenced by factors such as fuel load,wind, altitude, climb or descent and whether the birdsare flying in flock formation (Pennycuick 1969, 1975,1989, Tucker 1973, Hedenstrom and Alerstam 1995).To what extent do birds actually behave according tothese theoretical predictions, and what is their ability tofine-tune their flight speed in relation to the differentfactors? Although there are a number of studies indicat-ing that birds do indeed change their flight speeddepending on wind and altitude (cf. Hedenstrom andAlerstam 1995), there is still little critical informationon this, because of the uncertainties affecting quantita-tive predictions of flight speeds from the present aero-

dynamic theory for flapping flight in birds. Rather thantrying to interpret measurements of absolute flightspeeds in relation to theory, a more fruitful approach isto compare flight performance of the same species indifferent situations in a qualitative or semi-quantitativeway, to test if differences are in accordance with theo-retical expectation.

In this study we use this approach to compare theflight performance of Brent Geese Branta bernicla ber-nicla during spring and autumn migration. We madedetailed measurements of flight speeds at two localitieswhere Brent Geese leave the coast to fly across land insouthernmost Sweden during their spring and autumnmigratory passage, respectively. We test the predictions(i) that the birds have a higher airspeed but a lower rateof climb during spring than autumn migration becauseof their much larger fuel loads during the spring pas-sage and, more speculatively, (ii) that they fly faster

© JOURNAL OF AVIAN BIOLOGY

JOURNAL OF AVIAN BIOLOGY 31:2 (2000) 215

during spring when migration may be more time-se-lected than during autumn (Alerstam and Lindstrom1990, Hedenstrom and Alerstam 1995).

Prior to long migratory flights birds store fat andprotein to be used as fuel on their journeys (Lind-strom and Piersma 1993). The increase in body massis expected to affect flight performance in a numberof ways (Hedenstrom 1992). Aerodynamic theories ofbird flight (Pennycuick 1969, 1975, 1989, Tucker1973, Greenewalt 1975, Rayner 1979) predict that air-speed (the birds’ flight speed relative to the surround-ing air) will increase with increasing mass. Thepositive relationship between load (extra mass) andairspeed is expected for all characteristic flight speedsirrespective of whether the bird flies to minimize en-ergy per overall flight time (Vmp), minimize energycost per distance travelled (Vmr) or minimize overalltime spent on migration (Vmt) (Hedenstrom and Aler-stam 1995). Based on the expected relationship be-tween load and airspeed, Pennycuick (1978) predictedthat cruising airspeed of a migrating bird should de-crease during the flight as the bird gets lighter. Thiscould be observed by measuring flight speeds of thesame population of birds at two widely separatedplaces along the route, as suggested by Pennycuick(1978) and Rayner (1990), but so far no such studyhas been made.

In fact, the only available empirical evidence abouteffects of load on flight speed come from experimentswith trained birds and bats flying with artificiallyadded mass. These experiments showed a reduction inflight speed with extra load (Videler et al. 1988,Videler 1995, Hughes and Rayner 1991), which iscontrary to theoretical expectation. These results maybe a consequence of the experimental situation andarrangement of the added mass (Rayner 1995), or toa flight situation of travelling between food patches(cf. Hedenstrom and Alerstam 1995). However,Videler (1995) suggested that an inverse correlationbetween load and airspeed might hold also for birdson migration, which is at odds with all traditionalpredictions.

By focusing on the migratory flights of Brent Geesewith clearly different fuel loads on their spring andautumn passage, we will critically address this issue.Brent Geese conduct their migrations in long flightsbetween traditional staging areas. In spring they flydirectly from staging areas in the Wadden Sea, west-ern Europe, to staging sites in the White Sea area,north-western Russia. In autumn they use the oppo-site route, from the White Sea to the Wadden Sea.Southernmost Sweden is situated at the beginning ofthe 2500 km flight in spring, relatively close to (300–500 km) the Wadden Sea, and at the end of the flightin autumn. Thus, autumn birds passing Sweden willhave consumed more of their fuel stores than springbirds and therefore be considerably lighter. This is

also validated by spring and autumn body massesmeasured in the Wadden Sea (Ebbinge 1989).

The difference in body mass between spring andautumn might not be the only factor creating a dif-ference in airspeeds between the seasons. If optimiza-tion criteria differ between spring and autumnmigration we would expect an even larger differencein flight speeds than the one caused by body massalone. Migrating birds have traditionally been ex-pected to fly with speeds that minimize energy expen-diture per distance covered (Vmr). However, Alerstamand Lindstrom (1990) pointed out that in order tominimize overall time spent on migration, includingrefuelling at stopovers, birds should fly faster thanVmr. A stronger element of time-selection in springthan in autumn seems likely in arctic-nesting birdslike Brent Geese, which have to reach the breedingareas within certain time limits to be able to breedsuccessfully in the short arctic summer.

In order to make a critical evaluation of differencesin flight speeds between spring and autumn migrationwith respect to fuel load and, possibly, time-selection,we must control for a number of additional factors.Migrating birds are expected to adjust flight speed toprevailing winds by increasing airspeed in headwindsand reducing it in tailwinds (Pennycuick 1978).Sidewinds may also affect flight speeds and optimalspeeds are expected to be even higher in sidewinds,especially in strong sidewinds, than in due headwinds(Liechti et al. 1994). There is an expected trade-offbetween horizontal airspeed and vertical speed (climbrate), airspeeds decreasing with increasing climb rate(Hedenstrom and Alerstam 1992). Flying in flock for-mation is expected to be associated with an optimalairspeed that decreases with increasing flock size(Hummel 1983), although the reverse relationship be-tween airspeed and flock size has been observed inwaders (Noer 1979). Characteristic airspeeds increasewith increasing altitude due to the decrease in airdensity (Pennycuick 1975).

According to flight mechanical theory the relativepower margin, i.e. the difference between power thatcan be produced by the flight muscles and the powerrequired to fly, becomes reduced with increasing bodymass. This happens because power required to fly in-creases faster with body mass than does power avail-able from the muscles (Pennycuick 1969, 1989).Assuming that climbing capacity is limited by theavailable power margin, an increased load will reducethe maximum climbing capacity (Hedenstrom andAlerstam 1992). We will therefore analyse not onlyflight speeds but also vertical speeds of climbingflocks of Brent Geese, testing the prediction that theheavier spring birds show lower climb rates than thelighter autumn birds.

216 JOURNAL OF AVIAN BIOLOGY 31:2 (2000)

Methods

Predicted flight speeds were calculated according toPennycuick (1975, 1989: prog. 1), with modified assump-tions about profile power ratio (=8.4/AR, where AR isaspect ratio; Pennycuick 1995). Recent wind tunnelstudies have suggested that the coefficient of body drag(Cdb) is considerably lower than previously thought(Pennycuick et al. 1996). In our calculations we usedboth the original estimate of Cdb (0.25) and the suggestednew, lower one (0.10; Pennycuick et al. 1996). Data onwing span and body mass of Brent Geese were takenfrom Ebbinge and Spaans (1995) and Ebbinge (1989).Spring body mass was taken as the mean of mid May

body masses of all birds (both sexes). This value repre-sents a conservative estimate of departure mass sincedeparture from the Wadden Sea staging areas, for theflight over Sweden towards the Arctic, takes place in lateMay. Autumn body mass was taken as mean body massof birds in the Wadden Sea in October, just after arrivalfrom the flight over Sweden. As lean body mass we usedthe lowest monthly average mass during the year fromEbbinge (1989). For wing area we used the value fromEbbinge and Spaans (1995) corrected to include the bodyarea between the wings as defined by Pennycuick (1989).Predicted flight speeds were calculated for sea levelconditions with an air density of 1.23 kg m−3.

Fig. 1. Maps over north-western Europe showing main staging areas Wadden Sea and White Sea (a) and southernmost Sweden(b) showing study sites Lund, Vitemolla and Bjorka.


Table 1. Body mass, morphological data and predicted flightspeeds from flight mechanical theory (Pennycuick 1975, 1989)for Brent Geese on spring and autumn migration in SouthSweden. Aspect ratio is the ratio of wing span to mean chordof the wing, calculated as (wing span)2/(wing area). Predictedflight speeds are shown both for coefficient of body drag(Cdb)=0.25 (a) and for Cdb=0.10 (b).

AutumnSpring

Lean body mass (kg) 1.25 1.25Total body mass (kg) 1.60 1.35Wing span (m) 1.20 1.20Wing area (m2) 0.146 0.146Aspect ratio 9.9 9.9Vmp (m s−1) 12.0a 14.3b15.1b 11.3a

Vmr (m s−1) 18.7a 23.6b 17.8a 22.4b

away is 0.7° in elevation and azimuth and 12 m inrange (Hedenstrom and Alerstam 1992). To determineairspeeds of the flocks, wind direction and speed weremeasured by tracking hydrogen- or helium-filledweather balloons, released within an hour of bird track-ing. Airspeeds were then calculated by vector subtrac-tion of wind velocity at the altitude where the birdswere flying from the birds’ flight tracks (track directionand groundspeed). For climbing flocks flying throughseveral altitude strata, wind measurements from therelevant strata were used and airspeed calculated as theaverage of all airspeed calculations during the entiretracking.

For analyses of vertical speed (climb rate) we selectedsequences showing steady ascent of at least 0.2 m s−1

for more than 120 s. In spring several flocks passingLund were climbing very slowly, at rates intermediatebetween level flight and true climbing (climb ratesbetween 0 and 0.2 m s−1). These flocks have beenomitted from the analysis of climb rate since we wereonly interested in flocks expected to use maximumsustainable power during the climb. For flocks withseveral climbing phases during a tracking, or withintervals of different rates of ascent during the climb,we selected the sequence with the highest climb rate,provided it lasted at least 120 s. All trackings refer tovisually identified, migrating flocks. The radar used inautumn was equipped with 9× and 18× binocularsfor identification of radar targets. In spring, BrentGoose flocks were spotted and counted with 10×binoculars or a 30× telescope and subsequently lo-cated and tracked with the range finder or by radar.

Results

Predicted flight speeds of Brent Geese

Predicted characteristic airspeeds of Brent Geese areshown in Table 1. Here we show values for bothminimum power speed (Vmp) and for maximum rangespeed (Vmr). The main purpose of calculating predictedcharacteristic airspeeds is not to present their absolutevalues but the expected difference in speed betweenspring and autumn based on the body mass difference.The predicted differences due to mass are 0.7–0.9m s−1 with Cdb=0.25 and 0.8–1.2 m s−1 withCdb=0.10. Hence, we predict that the Brent Geese flyfaster by c. 1 m s−1 on their passage in spring than inautumn, if they adjust airspeed in relation to their massaccording to flight mechanical theory. Should there bea considerably larger difference in speed between springand autumn, this would indicate that the geese adapttheir flight speed in response to time-selection to ahigher degree in spring than in autumn because time-selected flight speeds are faster than energy-selectedspeeds as shown by Alerstam and Lindstrom (1990) andHedenstrom and Alerstam (1995, 1997).

Flight speeds of migrating Brent Geese were recordedmainly at two sites in Skane, South Sweden (Fig. 1).Spring migrating birds were tracked at Lund (55°42%N,13°12%E) by using an optical range finder (WILD, 80cm, 11.25× ) with azimuth and elevation scales, fromthe roof terrace of the Ecology Building (83.5 m asl) inlate May–early June of 1995 and 1996 and by using atracking radar (X-band, 200 kW peak power, pulseduration 0.25 ms, pulse repeat frequency 500 Hz, 1.5°pencil beam width, antenna 91.5 m asl) in late May–early June of 1997 and 1998. Autumn migrating birdswere tracked with a mobile tracking radar (X-band, 40kW peak power, pulse duration 0.3 ms, pulse repeatfrequency 1800 Hz, 2.2° pencil beam width) at Vite-molla (radar antenna 30 m asl) at the Baltic Sea coast(55°42%N, 14°12%E) in September and October of 1980–1982. In addition a few (4) radar trackings were madeat Bjorka (radar antenna 45 m asl), in South CentralSkane (55°39%N, 13°37%E) in October of 1986 and 1989.The topography of the different main localities aresomewhat similar with land rising from sea level toabout 100 m asl. On a smaller scale there is a smalldifference as land rises gently from sea level towardsLund while it rises quite steeply from sea level up to 30m asl at Vitemolla. Thereafter the relief of the land-scape is a gently rising one also at Vitemolla, compara-ble to the situation at Lund.

The optical range finder trackings were made atdistances of 300 m–3 km and fixes were normally readevery 20 s. Previous calibration measurements to fixedobjects at known distances indicate that the accuracy ofrange measurements was 910 m at 500 m, 920 m at1 km and 9100 m at 2 km (Hedenstrom and Alerstam1994).

The radar trackings were made at distances rangingfrom 300 m up to 30 km. All flocks passed initiallywithin 0.3–5 km, even the ones later tracked for verylong distances. Range, azimuth and elevation wererecorded by a computer every 2 s (Lund, spring), every10 s (autumn trackings in 1981–1989) or every 60 s(autumn trackings in 1980, n=14). Maximum trackingerror for a radar target in high speed motion 1 km


Tracking data

The general situation during which the trackings weremade was similar between spring and autumn. Inspring, the geese fly inland at Lommabukten west ofLund (Fig. 1) and the majority of flocks was eitherclimbing or in level flight. Some of the longer radartracks contained phases of both climb and level flight.In autumn, most tracks were of flocks flying inland atVitemolla (Fig. 1) either with an initial phase of climbfollowed by level flight or climbing throughout thetracking. In addition, some flocks in autumn weretracked in level flight low over the Baltic Sea (n=10)or at higher altitude over central Skane (n=4). Omit-ting the low altitude flocks flying over the Baltic Sea didnot change any of the results so they remain included inthe analysis.

Number and duration of tracks are shown in Table 2a and b for the complete sample of trackings and theclimbing sequences, respectively. Most tracks were of2–10 min duration (72% in both seasons); 13% (spring)and 28% (autumn) were of more than 10 min duration.Both data sets thus represent long continuous trackingsof migratory flight. The climbing sequences were ofsimilar duration in spring and autumn with the major-ity between 2 and 5 min.

Effect of wind, climb rate, flock size and flightaltitude on airspeed

In Table 3 a we present summary statistics for flightspeeds and factors that might affect flight speeds duringspring and autumn. Groundspeeds and airspeeds weresignificantly higher in spring than in autumn (ground-speed: t-test, pB0.001, df=132; airspeed: pB0.001,df=132).

Before considering the differences in flight speedsbetween spring and autumn, we will present the analy-ses of factors that might affect observed flight speeds.Wind situations differed slightly but significantly be-tween spring and autumn, with a higher proportion oftailwind tracks in spring and of headwind tracks inautumn (Table 3 a). In both seasons there was asignificant negative relationship between airspeed andwind effect (speed decrement/increment due to wind,measured as groundspeed minus airspeed; Table 4, Fig.2). The change in airspeed with wind effect was thesame in spring and autumn (Fig. 2).

Dividing the material in headwind and tailwind situa-tions (wind effect B0 resp. \0) and analysing theeffect of wind, revealed that the change in airspeedseemed to be confined to headwind situations only. Inheadwinds the geese increased airspeed with increasingwind speed whereas no significant relationship wasfound with increasing tailwinds. The regression equa-tions in headwinds were: Y=18.4−0.46 X (r=0.43,n=36, p=0.01) in spring, and Y=15.1−0.64 X (r=0.47, n=32, p=0.01) in autumn, where Y is airspeedand X is wind effect (groundspeed minus airspeed).

The regression coefficients indicate a somewhatgreater change of airspeed during autumn, but the 95%confidence intervals of the slopes overlapped broadly(−0.13 to −0.79 in spring vs −0.19 to −1.09 inautumn). The mean airspeed in tailwinds was 18.5 ms−1 (s.d. 2.8 m s−1, n=40) in spring and 16.8 m s−1

(s.d. 2.1 m s−1, n=26) in autumn. Mean airspeed inheadwinds was 19.5 m s−1 (s.d. 2.7 m s−1, n=36) inspring and 17.7 m s−1 (s.d. 3.2 m s−1, n=32) inautumn.

Strong sidewinds, with an angle between track andheading exceeding 30° (cf. Liechti et al. 1994) occurredon very few occasions during the trackings, on none in

Table 2. Summary of number of tracks, track duration and total tracking time for the total material (a) and for the selectedclimbing sequences (b).

a. Total material AutumnSpring

No. of tracks 42 0Optical range finderTracking radar 34 58

5876Total

355 523MeanTrack duration (s)Range 42–2120 120–1330

Total tracking time (s) 26 976 30 320

Autumnb. Climbing sequences Spring

0No. of tracks Optical range finder 1737Tracking radar 10

3727Total

272Track duration (s) Mean 276Range 120–1020120–1000

7465 10 064Total tracking time (s)


Table 3. Summary statistics of flight speeds and factors that might influence flight speeds of Brent Geese on spring and autumnmigration in South Sweden for the total material (a) and for selected climbing sequences (b). Wind effect is expressed asGroundspeed minus Airspeed. Differences between spring and autumn were tested with Student’s t-test for groundspeed,airspeed, wind effect, climb rate and with Mann-Whitney U-test for flock size and flight altitude. Levels of significance: *pB0.05, ** pB0.01, *** pB0.001, ns=non significant.

a. Total material Level of significanceSpring (n=76) Autumn (n=58) DifferenceMean (9s.d.) Mean (9s.d.)

Groundspeed (m s−1) 19.7 (3.9) ***16.0 (3.9) 3.7Airspeed (m s−1) ***19.0 (2.8) 17.3 (2.8) 1.7Wind effect (m s−1) 0.7 (3.8) **−1.3 (3.8) 2.0

nsClimb rate (m s−1) 0.11 (0.31) 0.14 (0.28) −0.03Flock size (no. of birds) 280 (278) 76 (84) 204 ***Flight altitude (m agl) 341 (163) ***215 (172) 126

b. Climbing sequences Spring (n=27) Autumn (n=37) Difference Level of significanceMean (9s.d.) Mean (9s.d.)

Groundspeed (m s−1) ***19.9 (4.6) 13.9 (3.4) 6.0Airspeed (m s−1) 18.4 (3.1) ***15.4 (2.8) 3.0Wind effect (m s−1) 1.4 (3.4) −1.5 (4.0) 2.9 ***Climb rate (m s−1) 0.46 (0.27) *0.62 (0.27) −0.16Flock size (no. of birds) 272 (254) ***60 (76) 212Flight altitude (m agl) 446 (199) ***241 (192) 205

Table 4. Coefficients of correlation between airspeed and some factors that might affect airspeeds for Brent Geese on spring andautumn migration in South Sweden. * pB0.05, ** pB0.01, ns=non significant.

Level of significanceSpring (n=76) Level of significance Autumn (n=58)

*Wind effect −0.31 ** −0.31Climb rate −0.22 nsns (p=0.06) −0.14Log Flock size −0.08 ns (p=0.06)ns 0.25Flight altitude 0.01 ns 0.26 *

spring and only four in the autumn. Thus we did notanalyse the possible effect of sidewinds any further.

Overall climb rates did not differ between spring andautumn for the total sample (Table 3 a). Airspeed wasnot significantly correlated with climb rate in any of thetwo seasons, although a tendency towards a negativerelationship was found in spring (Table 4). Flock sizeswere much larger in spring than in autumn (Table 3 a,Fig. 3). During both seasons most tracks were of flocksranging from several tens up to several hundreds ofbirds. In these intervals we do not expect any largeimpact of increasing flock size on airspeed. For aerody-namic reasons the largest effect should be found be-tween birds flying singly and in flocks of up to about10–15 birds (Lissaman and Shollenberger 1970, Hum-mel 1983). No significant correlations were found be-tween flock size and airspeed, but there was a tendencytowards increasing airspeed with increasing flock size inautumn (Table 4). Flight altitudes in spring were higherthan in autumn (Table 3 a, Fig. 4), but the majority oftracks was below 500 m where only a limited effect onairspeed is expected due to decreasing air density. Still,in autumn airspeed increased slightly with increasingaltitude (Table 4).

Fig. 2. Linear regressions of recorded airspeeds in relation towind effect (groundspeed minus airspeed) of Brent Geese onspring (open circles) and autumn (filled squares) migrationover southernmost Sweden. Regression equations are Y=19.1−0.23 X (r= −0.31, n=76, p=0.01) in spring (dottedline) and Y=17.0−0.23 X (r= −0.31, n=58, p=0.02) inautumn (solid line) where Y is airspeed and X is wind effect.


Fig. 3. Distribution of flock sizes in the total sample oftrackings of Brent Geese on spring and autumn migration insouthernmost Sweden. Arrows show means.

any of the other factors (wind effect, flock size, flightaltitude) had any significant effect on climb rates (Table5), although there was a tendency towards increasingclimb rates with increasing altitude in autumn (r=0.32,p=0.06).

Considering only the climbing flights, airspeedsshowed a significant negative correlation with windeffect in autumn but not in spring (Table 6). No effectof climb rate on airspeed was found, but averageairspeeds in climbing flight (18.4 m s−1 in spring and15.4 m s−1 in autumn) were slightly lower than overallmean airspeeds (19.0 m s−1 and 17.3 m s−1). Compar-ing airspeeds between climbing sequences and phases oflevel flight of the same flocks showed significantly lowerairspeed during climb than level flight in autumn(means: 14.2 m s−1 vs 18.7 m s−1; paired t-test,p=0.0001, df=19). In spring sample size was toosmall for a meaningful analysis. Flock size and flightaltitude showed no significant correlation with airspeedduring climb (Table 6).

Discussion

Difference in airspeeds

Even if airspeeds of migrating Brent Geese were quitevariable between flocks, ranging between 13 and 25 ms−1 (with a similar degree of variation in spring andautumn, Fig. 5), their mean airspeed was clearly fasterduring spring than autumn (Table 3 a, Fig. 5).Recorded airspeeds were close to predicted Vmr calcu-lated with Cdb=0.25 and fell between predicted Vmp

and Vmr calculated with Cdb=0.10 (Table 1, Table 3 a).The difference between mean spring and autumn air-speeds was 1.7 m s−1, i.e. about 10% of the overallmean. Airspeed was significantly correlated with windduring both spring and autumn migration, but thedifference between seasons remained irrespective of the

Flight speeds in spring and autumn

Groundspeed and airspeed were higher in spring thanin autumn (Table 3 a, Fig. 5 a and b). Part of thedifference in groundspeed was explained by a largerproportion of tailwind tracks during spring, but animportant part was also due to significantly differentairspeeds between the seasons. The overall difference inmean airspeeds was 1.7 m s−1. For the mean airspeedsin Table 3 a, the 95% confidence intervals were 18.3–19.6 m s−1 in spring and 16.6–18.0 m s−1 in autumn.The intercepts of the regression lines in Fig. 2 show thatthe average airspeeds at zero wind were 19.1 m s−1 inspring and 17.0 m s−1 in autumn, with 95% confidenceintervals of 18.5–19.7 m s−1 in spring and 16.2–17.7 ms−1 in autumn.

Climbing sequences

Vertical speeds during climbing flight were significantlylower in spring than in autumn (Table 3 b, Fig. 6). Thegeneral differences between spring and autumn in hori-zontal flight speeds (groundspeed and airspeed), windeffect, flock size and altitude were very similar to thosedescribed for the total material (Table 3 a, b), as theclimbing sequences make up a subsample of the totaltracking material. Neither horizontal flight speeds nor

Fig. 4. Distribution of flight altitudes in the total sample oftrackings of Brent Geese on spring and autumn migration insouthernmost Sweden. Arrows show means.


wind effect (Fig. 2) and was on average 2.1 m s−1 whenaccounting for the wind effect by calculating the air-speeds at zero wind (corresponding to the intercepts inFig. 2).

Vertical speed (climb/descend rates) did not differbetween the two seasonal total samples of tracks (Table3 a), so this factor cannot explain the difference be-tween spring and autumn airspeed. Average flock size

Fig. 6. Distribution of vertical flight speeds, climb rates, forselected climbing sequences of Brent Geese on spring andautumn migration in southernmost Sweden. Arrows showmeans.

Fig. 5. Distribution of groundspeeds (a) and airspeeds (b) ofthe total sample of Brent Geese on spring and autumn migra-tion in southernmost Sweden. Arrows show means.

and altitude, however, differed significantly between thetwo seasons (Table 3 a), but for two reasons it is highlyunlikely that the difference in airspeed was due to anyof these factors: (1) The theoretically expected effectswould be minor (affecting airspeed by less than 1%)compared with the observed seasonal difference in air-speed (cf. Lissaman and Shollenberger 1970, Hummel1983, Pennycuick 1975). By way of example, character-istic flight speeds are expected to increase by a mere 5%per 1000 m increase in altitude (calculated from Penny-cuick 1975) because of the reduction in air density. (2)There was no significant correlation between flock sizeand airspeed in either season, and only a weak correla-tion between altitude and airspeed in autumn (Table 4).

Differences in weather (barometric pressure) and agecomposition are two further factors with a potentialinfluence on airspeed, but we think that also thesefactors are highly unlikely to provide an explanationfor the observed seasonal difference in airspeed. Ifspring tracks were consistently recorded in low pressuresituations and autumn tracks in high pressure situationswe would expect airspeeds in spring to be higher thanin autumn. However, the difference in air density be-tween low and high pressures is relatively small andwould give differences in airspeeds of only about 1%(Pennycuick 1975, 1989). Furthermore, tracks in bothspring and autumn were made in varying weather con-ditions, with a dominance of high pressure situations inboth seasons.

The presence or absence of juvenile birds might affectairspeeds. Juvenile Brent Geese are structurally smallerand lighter than adult birds (Cramp 1977), and maytherefore fly more slowly than adults. Brent Geesemigrate family-wise during autumn and if adult birds


Table 5. Coefficients of correlation between climb rate and some factors that might affect climb rates for Brent Geese on springand autumn migration in South Sweden, selected climbing sequences. ns=non significant.

Spring (n=27) Level of significance Autumn (n=38) Level of significance

Airspeed −0.04 ns 0.05 nsns0.09ns0.30Wind effect

Log Flock size 0.10 ns 0.15 nsFlight altitude 0.01 ns 0.32 ns (p=0.06)

adjust airspeed to that of the smaller juveniles (which isunknown if they do) and if our autumn trackingsmainly consisted of families with young birds we wouldexpect lower airspeeds in autumn. We do not know theage composition of our tracked flocks, but since ourmaterial spans several different spring and autumnseasons, and the breeding success of Brent Geese ischaracterized by seasons with almost no fledged juve-niles alternating with seasons with up to 50% juvenilesapproximately every third year (Bergmann et al. 1994),we find it highly unlikely that we have tracked onlyfamily flocks with young. Furthermore, young birds arestill lighter than adults in spring but heavier than youngbirds in autumn (Cramp 1977), so any effect of youngbirds would affect recorded spring airspeeds as well.

We conclude that there was a significant difference inairspeed between spring and autumn that cannot beexplained by differences in wind situations, flock sizes,flight altitudes, weather (barometric pressure) or agecomposition of tracked flocks. The difference (2.1 ms−1, 0.8–3.1 m s−1 in a no wind situation) is reason-ably close to what we expect from flight mechanicaltheory based on average differences in body mass be-tween spring and autumn (0.7–1.2 m s−1). Our resultsare thus in agreement with the hypothesis that birdsincrease their airspeed with heavier fuel load accordingto traditional flight mechanical theory, but they are atvariance with Videler’s (1995) proposal that heavierbirds decrease airspeed and instead increase lift byincreasing wingbeat amplitude and inclination of bodyand tail, enlarging wing and tail area and increasingwingbeat frequency. The change in flight behaviour thatVideler describes may occur for example when raptorscarry food to their nests but probably not when migra-tory birds carry extra loads within their bodies as fatand protein. Furthermore, Videler’s (1995) argumentsare based on an assumed steep increase in flight costwith increasing speed. It is probably true that the flight

cost increases rather steeply at very high flight speeds,but wind tunnel studies have shown a relatively moder-ate increase within the speed intervals that birds actu-ally use for cruising flight (reviewed in Norberg 1990,1995). Thus, increasing airspeed with increasing massmight not be as costly as Videler assumes and, further-more, migrating birds presumably build up their flightmuscle mass before departing with heavy fuel loads(Pennycuick 1978, Piersma 1998).

The difference in airspeed between spring and au-tumn seemed to be slightly larger than predicted fromthe seasonal mass difference alone. Could this, in addi-tion to a mass-dependent effect, be explained by aninfluence of a larger element of time-selection duringspring as compared to autumn migration, according toour original hypothesis? We cannot rule out such anadditional effect of time-selection, but there is no con-vincing evidence in favour of it since the discrepancybetween the observed seasonal difference in airspeedand the predicted mass-dependent difference is notlarge enough to be statistically significant.

We consider it most likely that the seasonal differ-ence in airspeed that we have recorded for migratingBrent Geese in southernmost Sweden is associated withthe difference in load between the seasons, thusstrongly supporting the hypothesis of mass-dependentflight-speed adjustment. The possibility of a more time-selected flight speed during spring than autumn migra-tion must remain speculative.

Climbing performance

Observed maximum sustained climb rates in spring(0.46 m s−1) were significantly lower than in autumn(0.62 m s−1), as expected from the difference in averagebody mass between the seasons (Table 3 b). Our au-tumn climb rates were slightly higher than those previ-

Table 6. Coefficients of correlation between airspeed and factors that might influence airspeed for climbing sequences of BrentGeese on spring and autumn migration in South Sweden. *** pB0.001, ns=non significant.

Level of significanceSpring (n=27) Level of significance Autumn (n=38)

−0.54ns0.02Wind effect ***−0.04 ns ns0.05Climb rate−0.10 ns ns0.01Log flock size

Flight altitude 0.23 ns 0.28 ns (p=0.10)


ously reported based partly on the same material(0.53 m s−1; Hedenstrom and Alerstam 1992). Thediscrepancy is due to the fact that we used a shorterminimum tracking interval, 120 s, than Hedenstromand Alerstam (1992) who selected sequences lasting atleast 240 s. Several of the shorter trackings used inthis study showed high climb rates. In both seasonsthere were a few sequences of very high climb rates,exceeding 1 m s−1. In some of these cases therecorded climb rates might have been affected by ver-tical winds. If birds fly in rising air (thermals, slopewinds or lee waves) their climb rates will of courseincrease. We have no means of controlling for theoccurrence of vertical winds (cf. Hedenstrom andAlerstam 1992, 1994), but it is likely that such windsoccur under certain circumstances at both our track-ing sites. In spring the highest climb rates wererecorded during late morning to early midday in clearweather, when the birds might have flown throughareas of rising air in thermals and getting extra lift.In autumn both occasions with observed climb ratesover 1 m s−1 (Fig. 6), occurred in easterly winds,when the relief of the coast probably can create slopewinds at low altitudes (Hedenstrom and Alerstam1992).

The climbing performance of the Brent Geese inthis study may be compared with that of Light-belliedBrent Geese Branta bernicla hrota migrating fromstaging areas in Iceland across the Greenland ice-capaccording to the satellite telemetry study of Gud-mundsson et al. (1995). The Light-bellied Brents puton large fuel loads in Iceland to cover the flight overGreenland towards arctic Canada. Departure bodymasses are around or even exceed 2.0 kg (Gud-mundsson et al. 1995). Apart from the difference infuel loads these birds are morphologically very similarto the Brent Geese passing Sweden in spring and au-tumn (Cramp 1977, Gudmundsson et al. 1995). Birdsflying over the Greenland ice-cap were found to haveextremely low climb rates and slow forward flightspeeds indicating that they actually had to land andrest intermittently while climbing up the slope of theGreenland ice. It seems that body masses of about2.0 kg (54% extra load relative to lean body mass)are the largest Brent Geese can carry and still retaina (marginal) capacity to climb (Gudmundsson et al.1995). Brent Geese departing from the Wadden Seatowards Siberia also put on large amounts of fat andprotein (32%) but still, judging from this study, donot face any serious limitations of climbing capacity.These two studies in combination show that climbingcapacity may be limited when birds carry very largefuel loads (Greenland birds) but also that relativelylarge birds, such as Brent Geese, can put on substan-tial amounts of fat and protein without facing anyserious restrictions on climbing capacity (spring birdsin South Sweden).

Acknowledgements – Financial support for this study wasobtained from the Natural Science Research Council to T.A.,and from Gustaf Danielsson’s Foundation (Swedish Ornitho-logical Society) to M.G. Felix Liechti and A, ke Norberg gavevaluable comments on the manuscript.

ReferencesAlerstam, T. and Lindstrom, A, . 1990. Optimal bird migra-

tion: the relative importance of time, energy and safety.– In: Gwinner, E. (ed.). Bird Migration: Physiology andEcophysiology. Springer, Berlin pp. 331–351.

Bergmann, H.-H., Stock, M. and ten Thoren, B. 1994.Ringelganse. – AULA-Verlag, Wiesbaden.

Cramp, S. (ed.). 1977. Birds of the Western Palearctic. –Oxford University Press, Oxford.

Ebbinge, B. S. 1989. A multifactorial explanation for varia-tion in breeding performance of Brent Geese Branta ber-nicla. – Ibis 131: 196–204.

Ebbinge, B. S. and Spaans, B. 1995. The importance ofbody reserves accumulated in spring staging areas in thetemperate zone for breeding in Dark-bellied Brent GeeseBranta b. bernicla in the high Arctic. – J. Avian Biol. 26:105–113.

Greenewalt, C. H. 1975. The flight of birds. – Trans. Am.Phil. Soc. 65: 1–67.

Gudmundsson, G. A., Benvenuti, S., Alerstam, T., Papi, F.,Lilliendahl, K. and A, kesson, S. 1995. Examining the lim-its of flight and orientation performance: satellite track-ing of brent geese migrating across the Greenlandice-cap. – Proc. R. Soc. Lond. B 261: 73–79.

Hedenstrom, A. 1992. Flight performance in relation to fuelload in birds. – J. theor. Biol. 158: 535–537.

Hedenstrom, A. and Alerstam, T. 1992. Climbing perfor-mance of migrating birds as a basis for estimating limitsfor fuel-carrying capacity and muscle work. – J. Exp.Biol. 164: 19–38.

Hedenstrom, A. and Alerstam, T. 1994. Optimal climbingflight in migrating birds: predictions and observations ofknots and turnstones. – Anim. Behav. 48: 47–54.

Hedenstrom, A. and Alerstam, T. 1995. Optimal flightspeeds in birds. – Phil. Trans. R. Soc. Lond. B 348:471–487.

Hedenstrom, A. and Alerstam, T. 1997. Optimum fuel loadsin migratory birds: distinguishing between time and en-ergy minimization. – J. theor. Biol. 189: 227–234.

Hughes, P. M. and Rayner, J. M. V. 1991. Addition ofartificial loads to long-eared bats Plecotus auritus : handi-capping flight performance. – J. Exp. Biol. 161: 285–298.

Hummel, D. 1983. Aerodynamic aspects of formation flightin birds. – J. theor. Biol. 104: 321–347.

Liechti, F., Hedenstrom, A. and Alerstam, T. 1994. Effectsof sidewinds on optimal flight speed of birds. – J. theor.Biol. 170: 219–225.

Lindstrom, A, . and Piersma, T. 1993. Mass changes in mi-grating birds: the evidence for fat and protein storagere-examined. – Ibis 135: 70–78.

Lissaman, P. B. S. and Shollenberger, C. 1970. Formationflight of birds. – Science 168: 1003–1005.

Noer, H. 1979. Speeds of migrating waders Charadriidae. –Dansk Orn. Foren. Tidskr. 73: 215–224.

Norberg, U. M. 1990. Vertebrate Flight. – Springer, Berlin.Norberg, U. M. 1995. Energetics of flight. – In: Carey, C.

(ed.). Avian Energetics and Nutritional Ecology. Chap-man & Hall, New York, pp. 199–249.

Pennycuick, C. J. 1969. The mechanics of bird migration. –Ibis 111: 525–556.

Pennycuick, C. J. 1975. Mechanics of flight. – In: Farner,D. S. and King, J. R. (eds). Avian Biology. Vol 5. Aca-demic Press, New York, pp. 1–75.


Pennycuick, C. J. 1978. Fifteen testable predictions about birdflight. – Oikos 30: 165–176.

Pennycuick, C. J. 1989. Bird Flight Performance: a PracticalCalculation Manual. – Oxford University Press, Oxford.

Pennycuick, 1995. The use and misuse of mathematical flightmodels. – Israel J. Zool. 41: 307–319.

Pennycuick, C. J., Klaasen, M., Kvist, A. and Lindstrom, A, .1996. Wingbeat frequency and the body drag anomaly:Wind-tunnel observations on a thrush nightingale (Luscinialuscinia) and a teal (Anas crecca). – J. Exp. Biol. 199:2757–2765.

Piersma, T. 1998. Phenotypic flexibility during migration: opti-mization of organ size contingent on the risks and rewardsof fuelling and flight? – J. Avian Biol. 29: 511–520.

Rayner, J. M. V. 1979. A new approach to animal flightmechanics. – J. Exp. Biol. 80: 17–54.

Rayner, J. M. V. 1990. The mechanics of flight and birdmigration performance. – In: Gwinner, E. (ed.). Bird Migra-tion: Physiology and Ecophysiology. Springer, Berlin, pp.283–299.

Rayner, J. M. V. 1995. Flight mechanics and constraints onflight performance. – Israel J. Zool. 41: 321–342.

Tucker, V. A. 1973. Bird metabolism during flight: evaluationof a theory. – J. Exp. Biol. 58: 689–709.

Videler, J. J. 1995. Consequences of weight decrease on flightperformance during migration. – Israel J. Zool. 41: 343–353.

Videler, J. J., Groenewegen, A., Gnodde, M. and Vossebelt, G.1988. Indoor flight experiments with trained kestrels. Theeffect of added weight on flapping flight kinematics. – J.Exp. Biol. 134: 173–183.

(Recei6ed 5 January 1999, accepted 18 March 1999.)


flight speeds and climb rates of brent geese: mass-dependent differences between spring and autumn...

Documents